Why reain_graph is needed here ? #96

Open

opened

Hi, I read the GAN code and I have the confusion that why the retain_graph is needed here ?

PyTorch-Tutorial/tutorial-contents/406_GAN.py

Line 71 in 4e5122c

D_loss.backward(retain_graph=True) # reusing computational graph

The computation graph of discriminator should not be released ?

Metadata

Assignees

No one assigned

Labels

No labels

No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests